Pink House Technology · Partner packet · 19 October 2026
I have a business concept. It is not an app store listing and it is not a warehouse. It is an engine.
The engine is a database of aesthetic attributes. Hair color. Face shape. Body type. Hues. Styles. Those are data points. They feed an AI that generates designs that sit next to real intellectual property without being that property. Near-IP-adjacent. Non-infringing. Interesting. Cute. Something a person would actually wear or post.
The pictures go out free on TikTok and YouTube. If anything wants to become a shirt, it goes to TeePublic as print-on-demand. No SKUs of my own. No inventory. No failures in the old sense, because nothing is committed until it works. If it does not work, it was an idea on a screen. If it works, the printer prints it after someone already wanted it.
That is the commerce. I am calling the larger thought Commerce in the 30s. I already have a supply-chain book on Pinkhouse called Fine Tuning DTC for Profit 2027. That book is the other half of my life: freight, HTS, Mexico, de minimis dying on 29 August 2025. This packet is not that book. This packet is the aesthetic engine, the fandoms, the science of picking data points, the ask, and the kind of people who can build it. White hat. Something nobody else has ever done.
I do not have the skill sets or the patience to build the database and do the nuance myself. I need people who walk in and say we can work that out. Positive people. People smart enough to get it and do it. People who can make cheap mistakes. If you like the vision, you fund the data. Vision first. Data second. Execution third. That is the ask I would put in front of someone like Jack Ma. Not “buy my warehouse.” Buy the idea hard enough to pay for the library underneath it.
Start with what a human already does when they love a look. They do not love a trademark. They love a cluster. A haircut. A color that belongs to an era. A way a jacket sits on a shoulder. A face shape that the culture has decided is this year’s pretty. A body type the posters keep repeating. A hue that means a comeback stage, or a third kit, or a night race.
Those things can be named. If they can be named they can be stored. If they can be stored they can be combined without copying a person or a logo.
The database is not a mood board on a laptop. It is attributes. Hair color as a field. Face shape as a field. Body type as a field. Hue as a field. Style as a field. You can add more later: era, silhouette, fabric language, pose energy, “cute vs sharp,” generation of a wave. The point is a structured library of how a look is built, not a folder of stolen stills.
The AI reads that library and generates. The output should feel like it came from the same weather system as the original IP. Same pressure. Same season. Not the same face. Not the same trademark. A person who loves the original should feel recognized. A lawyer for the original should not have a clean shot.
That is the hard line. Near-IP-adjacent. Non-infringing. I mean both words. Adjacent is the product. Non-infringing is the license to stay in business. White hat is not a slogan here. It is the whole shop. If we cannot do it clean, we do not do it.
K-pop is the lead example because K-pop already is a database wearing human beings. Idols are catalogued by hair, era, concept, comeback palette, outfit architecture. Fans argue about the look. The look changes on a schedule. There is always a next wave. The audience is huge. The visual bible is distinct. That is the shape. If K-pop is not number one for this engine, whatever wins still has to have that shape: a look bible, a fighting crowd, and yearly waves.
Here is how I want a partner to see the first pair of pictures.
On one side: a real IP piece. Something that already exists in the world, that fans already love, that is cute and interesting. You are allowed to look at it. You are not allowed to reprint it as if you own it.
On the other side: an original adjacent piece. Same heat. Same cuteness. Same “I would screenshot that.” Hair, face, body, hues, styles pulled from the attribute engine, not from a ripped photocall. A person who lives in that fandom should smile. A person who owns the group should not be able to say you stole the member.
Both pieces interesting. Both cute. That is the demo. If the original is magnetic and the adjacent is dead, the engine failed. If the adjacent is a clone, the engine failed in the other direction. The win is two things that belong in the same room and are not the same object.
K-pop is built for this. Hair color is a comeback announcement. Face shape is already a fan taxonomy. Body type and styling are part of how generations are argued. Hues come in concepts: dark, candy, military, school, cosmic, quiet luxury, whatever this year named. Styles shift in waves. A group’s 2016 look and 2026 look are not the same species. The database has to hold that motion. Not one frozen pretty. Waves.
Free on TikTok and YouTube is how the adjacent piece meets the crowd that already fights about the real one. TeePublic is how a winner becomes a shirt without me buying inventory. No SKUs in my closet. Ideas until the idea earns a print.
I said if K-pop is not the number one thing for this engine, what would be. The answer I gave on the call is: anything with the same shape. Huge interested audience. Distinct visual database. Waves of shifting styles. Then we listed them. They are not random merch categories. They are other look bibles.
Penguins. Sharks. I said those out loud because the engine is not only human faces. An aesthetic wave can be an animal the internet has already decided is a character type. Penguin has a silhouette, a tuxedo joke, a climate-era tenderness, a cute that sells. Shark has a different silhouette, a different humor, a different kind of tough. You still do attributes. You still do adjacent, not a studio mascot. You still do not print someone else’s cartoon.
Then the list that has the same shape as K-pop:
Anime visual grammar. Not Naruto. Not a studio. Hair physics, eye language, school-uniform silhouette, season palettes. Waves: shonen grit, shojo sparkle, nineties cel, current gloss. You sell rainy rooftop magenta bangs. You never sell a character.
Harajuku and J-fashion tribes. Lolita, gyaru, visual kei, Decora. The database is hair, face, body, hue. The tribes are already attributes. Waves every few years. Adjacent is the tribe, not a brand and not an idol you do not own.
Football kits and ultras. Home, away, third. Collar, stripe, scarf hue. Every season is a kit drop. No crests. No player names. Claret vertical, away mint collar. The crowd already fights over kits. That is the fighting crowd.
F1 liveries. Nose, halo, accent stripe, night-race neon. New paint every year. Papaya-speed block without a team wordmark. Motorsport already trains the eye on color blocks. The engine just stores them honestly.
Sneaker and streetwear seasons. Silhouette, panel, gum sole, collab colorway. Drops are the waves. Three-panel Chicago-ish, no swoosh. The audience is enormous. The look bible is already SKU language in the real world. We do not become a fake Nike. We become a library of how a colorway feels.
Pop-album eras. The Swiftie machine without the singer. A palette plus a font mood plus a tour silhouette. Each record cycle is a wave. Midnights navy fog as a hoodie idea. No lyrics. No face. The fandom already lives in eras. We live in attributes.
NBA city editions. City hue, skyline abstract, alternate jersey cut. Yearly waves. Harbor teal split, no franchise mark. Same as kits. Same as liveries. Civic color as a product language.
Cottagecore, dark academia, goblincore. Already an attribute language: tweed, moss, candle, round specs. TikTok renames the mood and that is a wave. Soft original illustration. Not a movie still.
Metal and goth subgenres. Black metal frost, hair metal chrome, nu-metal cargo, witch-house purple. Decades you can remix. Sigils we invent. Not band marks. The crowd fights. The look bible is old and still moving.
Game-world palettes belong in the same family: cyberpunk rain, solarpunk leaf-chrome, cozy-farm warmth. Waves follow whatever is loud this year. Neon alley umber. Never a Nintendo silhouette. Never a boss from a game we do not own.
K-pop is still the cleanest because the idols are the database. Next cleanest is anime grammar and kit-and-livery culture. Same hunger. Same shifting look. No trademark required if we stay adjacent and white hat.
Anything with a look bible, a fighting crowd, and yearly waves can run through the same engine. That sentence is the filter. If it does not have those three, it is not a lead vertical. It might be a side print. It is not the engine’s job.
I am not building a shop that guesses sizes and hopes. I have done commerce with trucks and boxes. I know what committed inventory feels like when the idea was wrong.
This concept does the opposite.
The designs go out free on TikTok and YouTube. That is distribution. That is also research. The crowd tells you, in public, what they will fight over. Views, saves, comments, stitches. The fighting crowd is not a problem. The fighting crowd is the market speaking.
TeePublic is print-on-demand. Someone likes the idea enough to wear it. Then it exists as cloth. Not before. No SKUs sitting in a garage. No “we printed ten thousand and they are in a unit in Mississauga.” No failures of that kind, because failure used to mean leftover units. Here leftover units do not get born.
I said no SKUs, just ideas. I mean the company does not live as a catalog of committed styles. The company lives as a generator of adjacent ideas. The ones that catch become products because a platform already knows how to print a shirt when a stranger pays. We do not have to become a 3PL to find out if the penguin was cute.
That is DTC without the old oxygen of de minimis, and without pretending we are Shein. We are not flooding America with five-dollar accidents. We are putting pictures in front of people who already have a look bible in their head, and letting the ones that sparkle become a print.
This is not “ask a model for a cute K-pop girl.” That is slop. YouTube is full of that. I have been watching AI personas. A woman, twenty-five to thirty, geopolitics, built by a team, or built by a prompt pipeline that looks like a team. That is a different business. This is not that.
The science is evocative selection and evaluation of data points.
Evocative means the data point has to wake something. Hair color is not a hex only. It is a hex that means a comeback, or a tribe, or a kit year. Face shape is not a cartoon slider. It is a way fans already sort beauty. If the field does not evoke, it does not belong in the library, or it belongs with a weaker weight.
Selection means we do not vacuum the internet and call it a database. We choose. White hat choosing. We do not scrape a member’s face into a latent soup and hope the lawyer is sleepy. We name attributes at a level that is culture, not identity of a person.
Evaluation means we score whether the adjacent piece did the job. Did it sit beside the real IP and still feel alive. Did it stay non-infringing. Did the crowd respond. Did the TeePublic idea move. That is evaluation of data points after they have been through the generator. Database marketing, done in a way I do not believe has been done before.
Old database marketing was: buy a list, mail the list, count the checks. This is: build a living library of how a culture looks, generate adjacent objects, let a fighting fandom vote with attention, print only what already won. The list is not names and addresses. The list is attributes. The campaign is a picture that could not have existed without that library. The conversion is a shirt nobody had to inventory.
I have spent a lot of my life on quality as conformance to requirement. Philip Crosby. The work did the job you named. 1929 in another piece of mine is a human book, guaranteed AI-free, a certificate that a person wrote every line. Crosby’s own book is 1979, Quality Is Free. I tell you that so you know I am not sloppy about whose idea is whose. Same here. The adjacent piece must conform to a requirement we name: evocative, non-infringing, cute enough to post, clean enough to sell. If we cannot name done, the model will try for a hundred percent of the wrong thing. IGLI is my line, not Crosby’s: sometimes you take the job to a clean sixty or eighty and you leave the grease. We still name the requirement. We do not swallow the whole internet in one gulp and call it a brand.
That is the science I want funded. Not a vibe. A library with rules, a generator with a white-hat fence, and an evaluation loop that is actually marketing, not a dashboard nobody believes.
I called it Commerce in the 30s on this call.
I already have, on pinkhouse.tech, Fine Tuning DTC for Profit, also titled with 2027, working title The 2027 DTC Merchandising Guide. That is the operations playbook after the United States killed the eight-hundred-dollar de minimis exemption. China-as-gravity is over. Mexico, velocity, HTS, 3PL. It says on the page it is not an AI product. It is the supply-chain work from before this shop had a nickname. Cover still coming. The page is the book, not always the full manuscript.
Commerce in the 30s is the name I want for the bigger thought: what selling looks like when the thirties of this century are the operating system. Attention first. Print after. Data as a look bible. AI as a generator inside a fence. No warehouse of unsold cleverness. The DTC book on Pinkhouse is proof I have done the unglamorous half. This concept is the other half: demand as aesthetics, not as leftover cartons.
A partner should not confuse the two files. One is how a North American brand survives the new duty math. One is how a new kind of merch company never buys the duty math until a human already chose the picture.
Vision first. Data second. Execution third.
If you like the idea, you fund the data.
That is the ask I would put to Jack Ma, or to anyone who has actually built a market at that scale. I am not walking in with a warehouse tour. I am walking in with a vision of an engine that has not existed: aesthetic attributes as a serious database, adjacent generation as a clean product, fandom as the channel, print-on-demand as the factory that appears only when needed.
Vision first means we agree what “adjacent and cute and legal” is. We agree K-pop can lead and the other waves are the same animal. We agree white hat. We agree no SKUs as a religion until the idea earns a unit.
Data second means the money goes into the library. People who can sit with hair, face, body, hue, style and not get romantic or sloppy. People who can encode a wave without encoding a person we do not have the right to encode. That is the expensive part. That is also the only part that makes the engine real. Without the data, we are another prompt account on TikTok.
Execution third means the generator, the posting, the TeePublic pipe, the evaluation loop. Execution without data is slop. Data without vision is a spreadsheet. Vision without either is a conversation I already had on a phone.
If you like the idea, you fund the data. If you do not like the idea, there is nothing to fund. I am not asking someone to fund my patience. I am asking someone to fund the library I cannot build alone.
Positive people. Avoid negative ones.
I mean it as an operating rule, not a mood. This work will get lawyer-adjacent. It will get fandom-adjacent. It will get “is this allowed.” Negative people turn that into a reason to never ship. Positive people turn that into a fence we can actually build.
People smart enough to get it and do it. Getting it is not enough. Doing it is the database. Doing it is the nuance. Doing it is looking at a generated piece and saying that hair is a person, pull it back to an attribute. Doing it is looking at a kit idea and saying that is a crest, kill it.
People who can make cheap mistakes. I do not need heroes who only swing at the IPO. I need a shop that can be wrong on a Tuesday on TikTok and not have forty thousand shirts. Cheap mistakes are the point of no inventory. If a person cannot stand being wrong in public for a day, they will force us back into SKUs so they can feel safe. That is the old commerce. We are not doing the old commerce.
I am honest about my limits. I do not have the skill sets to build this database. I do not have the patience to sit inside the nuance for a year and tag hues. I need people who walk in and say we can work that out. That sentence is the hiring test. If they need me to already have the ontology finished, they are not the people. If they hear the engine and start drawing the tables, they might be.
White hat required. Not later. Now. No stolen faces. No deepfake of a journalist. No “the fans won’t notice.” I have watched what AI YouTube is in October 2026. Synthetic experts. War-comment farms. The same avatar on thirty finance channels. That is not this company. We can use the same kinds of tools. We cannot use that ethic.
The goal is something nobody else has ever done. I know how that sounds. I still mean it. Plenty of people generate pretty girls. Plenty of people sell POD. Plenty of people love K-pop. The thing I do not see is a serious aesthetic-attribute database, evaluated like marketing science, fenced for non-infringement, aimed at fandom waves, with free attention as the factory gate and print only after. If someone has done that, I have not seen it. If we do it, it has to be clean enough to show Jack Ma without looking at our shoes.
A partner should hear a picture of done, because Crosby would ask, and because I get angry at that question and then admit it was the right question.
Done for a first season is not a hundred SKUs. Done is: the attribute library exists for one lead wave, almost certainly K-pop, with enough fields that a generator can make a pair of pictures — real-world reference on one side, original adjacent on the other — and a stranger in the fandom says both are interesting and cute, and a conservative IP read says the adjacent is not the member and not the mark.
Done is: those adjacent pieces post free on TikTok and YouTube on a cadence humans can stand behind, not a thirty-a-day farm.
Done is: a handful of ideas live on TeePublic as print-on-demand, no warehouse, and we can say which attributes correlated with attention. That is the science loop starting. Not a vanity chart. A sentence like: this hue plus this silhouette moved; that face-shape cluster got us in trouble; we killed it.
Done is: one other wave queued with the same engine — kits, or anime grammar, or penguins if the cute wants an animal — so we prove the engine is not a K-pop one-off.
Done is not: we became a media network of fake women explaining Iran. Done is not: we cloned a group and dared them to sue. Done is not: I sat alone and tagged ten thousand rows because nobody walked in.
It is not TokenScope. TokenScope is what you pay models. This is not a cost dashboard.
It is not Michelangelo. Michelangelo is governance before an output leaves a regulated system. This still needs a fence. It is not a pharma control point.
It is not an AI geopolitics presenter. I watched that. It is a team or a pipeline putting a face on commentary. Different hunger. Different risk. Different YouTube problem.
It is not Shein. We are not the firehose that killed de minimis for everyone else.
It is not a Six Sigma workbook. Quality here is conformance to the requirement we named: adjacent, non-infringing, evocative, cheap to be wrong, expensive only in data.
It is not Harry Crosby 1929. Different Crosby. I mention that only because I have been precise about years this week and I will stay precise. Philip Crosby is quality. The 1929 book in another article is a human-written object. This packet is commerce in the nineteen-thirties of the two-thousands, if you need the joke of the title spelled out. The thirties we are walking into.
The engine is a database of aesthetic attributes. Hair color, face shape, body type, hues, styles. It feeds an AI that generates near-IP-adjacent but non-infringing designs.
K-pop is the lead. A real IP piece beside an original adjacent piece. Both interesting. Both cute.
Other concepts if K-pop is not number one: penguins, sharks, anime visual grammar, Harajuku tribes, football kits, F1 liveries, sneaker drops, pop-album eras, NBA city editions, cottagecore, metal subgenres, game-world palettes. Anything with a look bible, a fighting crowd, and yearly waves.
Free on TikTok and YouTube. Print-on-demand on TeePublic. No SKUs. No inventory. No failures because nothing is committed until it works.
The science is evocative selection and evaluation of data points. Database marketing done in a way that has not been done before.
The book thought is Commerce in the 30s. The existing Pinkhouse book is Fine Tuning DTC for Profit 2027.
The ask: vision first, data second, execution third. If you like the idea, you fund the data.
The team: positive, not negative; smart enough to get it and do it; cheap mistakes. I cannot build the database or the nuance alone. I need people who say we can work that out.
White hat required.
The goal: something nobody else has ever done.
End of packet. Captured from the 19 October 2026 call. Phil Cheevers, Pink House Technology. philcheevers@pinkhouse.com